Terminal Market
Market Landscape
How observed market value is distributed across the modelling population, and where the model's held-out errors grow or lean.
- Population
- 5,820 players · 14 European first divisions
- Selection
- ≥ 450 club minutes in two years · 2025–26 valuations
- Errors
- held-out, 5-fold cross-validation · log scale
- Not
- the global market: a selected population
Observed value, by tier
players · share of 5,820Nearly half the population is valued under €2M; 345 players (5.9%) reach €30M. Tier boundaries are the project's fixed €2M / €10M / €30M cut-offs.
Where the model leans
mean residual by observed tier · log units- Emergingn 2,776−0.155B −0.496
- Establishedn 1,856+0.103B +0.300
- Eliten 843+0.232B +0.710
- World Classn 345+0.229B +0.662
Grouped by observed value, residuals pull toward the middle: Emerging players' values sit below Model A's estimates on average (−0.155), Elite players' above them (+0.232). Any regression shows this when results are grouped by the target itself; it describes the model, not mispriced players. Model B (dashed) leans further.
Positions
composition and held-out errorPlayers
Mean absolute error, log
- Goalkeeper0.459B 0.741
- Defender0.398B 0.726
- Midfield0.394B 0.718
- Attack0.413B 0.728
Model A Model B
Minutes and accuracy
held-out MAE by club minutes in two years · log- 450-899 minn 6920.521B 0.854
- 900-1799 minn 1,2230.430B 0.760
- 1800+ minn 3,9050.378B 0.692
Model A Model B
Less football, wider misses: Model A's typical error falls from ×1.68 for players with under 900 minutes to ×1.46 for regulars. Overall held-out MAE: 0.406 (A), 0.725 (B).
Player-level views
5,820 players in the local exportObserved value against age
Model A estimate against observed value
Dashed diagonal: estimate equals observed value. Above it the model is higher than the market.